High USP6NL Levels in Breast Cancer Sustain Chronic AKT Phosphorylation and GLUT1 Stability Fueling Aerobic Glycolysis

A large fraction of breast cancers, especially of the basal-like subtype, display unique gene ontologies of growth factor signaling, glycolysis and gluconeogenesis (1) and are characterized by increased glycolysis and dependence on this metabolic drift for proliferation (2, 3).

High rates of glucose uptake are achieved through transcriptional upregulation and increased targeting of GLUT1, a member of the glucose transporter family, to the plasma membrane (PM; ref. 4). Localization of GLUT1 to the cell surface is largely regulated through its endocytic traffic (5–11), controlled, in nontransformed cells, by cytokine stimulation (12, 13) or energy stress (6). In cancer cells, several oncogenic lesions, including hyper activation of the EGFR/PI3K/AKT signaling pathway, induce accumulation of GLUT1 at the cell surface, resulting in increased aerobic glycolysis (14–16). However, the molecular determinants of this signaling-metabolic rewiring are largely unknown.

USP6NL is a GTPase-activating protein (GAP), which stimulates GTP hydrolysis on various members of the Rab family, thereby inhibiting endocytosis of PM receptors and retrograde Golgi transport (17–20). We found that USP6NL is amplified and overexpressed in a subset of breast cancers, mainly belonging to the basal-like Integrative Cluster 10 subtype. USP6NL overexpression correlates with worse prognosis, suggesting a role in the natural history of the tumor. This is strengthened by data showing that functional ablation of USP6NL inhibits proliferation and glycolysis exclusively in breast cancer cells harboring high levels of this protein.

Our findings argue, therefore, that USP6NL is a regulator of the crosstalk between signaling and metabolic pathways whose elevation causes glycolysis addiction in breast cancer cells.

HCC70, HCC1187, H1299 (ATCC) were grown in RPMI (Sigma) with 10% FBS (Euroclone) and 1% glutamine. MCF10A (a gift of J.S. Brugge, Harvard Medical School, Boston, MA) were grown in DMEM/Ham's F-12 with EGF 20 ng/mL (Peprotech), 5% horse serum, hydrocortisone 0.5 mg/mL, cholera toxin 100 ng/mL, and insulin 10 μg/mL (Sigma). HMEC (hTERT-HME1) were from ATCC and grown in DMEM/Ham's F-12 with EGF 20 ng/mL, hydrocortisone 0.5 mg/mL, insulin 10 μg/mL, and 10% FBS. The genetic identity of the cell lines was confirmed by short tandem repeat (STR) profiling (Cell ID; Promega), finally repeated in July 2017. Cells were periodically tested for mycoplasma with Venor GM Kit (Minerva Biolabs).

For functional ablation experiments, equal number of cells were plated in presence of siRNA oligos, and a second round of transfection was repeated the day after on the adherent cells. Cells were harvested 48 hours later for biochemical experiments or counted for the “proliferation” experiments (see “cell counts,” below). Ablation of USP6NL was initially performed with three different oligos (Silencer Select oligos, Ambion, 50 pmol; negative control: siRNA#2) using Lipofectamine RNAiMAX (Invitrogen). Because the three oligos yielded comparable results, oligo s18721 alone was subsequently used, after confirming its specificity in rescue experiments.

Stable, inducible HCC70 and HMEC cell populations were generated by infection with tetracycline-inducible lentiviral vectors (pSLIK) carrying HA-tagged USP6NL or HA-USP6NLR150 (R150; refs. 17, 19). In these constructs, the sequence targeted by oligo s18721 was mutated with the Phusion Site-Directed Mutagenesis Kit (Thermo Scientific) in a silencing-resistant version: 5′-TAGACAGTATAATCACGCA-3′. Stable populations, including MCF10A-pSLIK-RAB5A (a gift from Giorgio Scita, IFOM, The FIRC Institute for Molecular Oncology Foundation, Milan, Italy; ref. 21), were established by neomycin selection and grown in media supplemented with 10% Tet system-approved FBS (Clontech). Expression of USP6NL, R150, or RAB5 was induced by doxycycline (1 μg/mL; Sigma) for 48 hours. HCC70 cells stably expressing PI3KCA were generated by infection with the retroviral construct pBABE puro Myr PIK3CA, a gift from Jean Zhao (Dana-Farber Cancer Institute, Harvard Medical School, Boston, MA; Addgene plasmid #12523). pLPCX-EGFP-myc-GLUT1 (22) was kindly provided by Jeffrey Pessin (The Albert Einstein College of Medicine, Bronx, New York) and transiently transfected using the X-tremeGENE reagent (Sigma).

Antibodies were: affinity-purified rabbit polyclonal anti-USP6NL (homemade); anti-AKT (2920), anti-phosphoAKT-Ser473 (4060), anti-phosphoAKT-Thr308 (2965), anti-phosphoGSK3β-Ser9 (9336), anti-phosphoEGFR (3777), anti-phosphoMET (3077) (Cell Signaling Technology); anti-GSK3β (E-11 sc377213), anti-HA (Y-11 sc-805; F-7 sc-7392), anti-HSP90 (sc-13119), anti-EEA1 (N-19 sc-6415), anti-EGFR (1005 sc-03), anti-Rab5 (D-11 sc46692), anti-MET (D-4 sc514148) (Santa Cruz); anti-GLUT1 (ab15309), anti-GAPDH (6C5 ab8245), anti-myc (9E10 ab32) (AbCam); anti-EGFR (LA22) (Millipore). Alexa Fluor Phalloidin 647 (A22287) was from Thermo Scientific. EGF and HGF were from R&D, MK2206 was from Selleckchem, bafilomycin from Sigma, and MG132 from Calbiochem.

IHC was performed on a tissue microarray (TMA) of 1,608 consecutive breast cancers and 99 normal samples from the European Institute of Oncology in Milan, Italy (years 1997–2000). Cases were annotated for clinical and pathological parameters and censored at 10 years from surgery. All tissues were collected via standard operating procedures approved by the Institutional Ethical Board, and informed consent was obtained for all tissue specimens linked with clinical data.

TMAs were stained in a Bond Max Automated IHC Vision Biosystem (Leica Microsystems) using the Bond Polymer Refine Detection Kit (DS9800). Three-micrometer-thick sections were prepared from formalin-fixed paraffin-embedded blocks, deparaffinized, pretreated with the Epitope Retrieval Solution 1 (pH 6) at 100°C for 20 minutes and then incubated for 30 minutes with anti-USP6NL (Assay Biotech C19466), 1:100 diluted in Bond Primary Antibody Diluent (AR9352) at final concentration of 10 μg/mL. Slides were acquired with the Aperio ScanScope system (Leica Microsystems).

Sample extraction, processing, and analysis of the METABRIC samples are described in ref. 23. Differential expression was determined using the limma package (24), followed by survival distributions using the log-rank test. Comparison between expression in the Integrative Cluster 10 and the rest of the METABRIC tumors was by two-sided t test. Survival analysis was by Kaplan–Meier estimates; associations with diseases specific survival was determined by the log-rank test for single factors or Cox multiple regression for multiple several factors. The Cox model was stratified by hospital and included tumor size, tumor grade, number of lymph nodes as numeric and the USP6NL copy number aberration as a binary variable. The increasing trend between copy number and expression was determined using the one-sided Jonckheere–Terpstra test from the R clinfun package. Statistical analysis in the cell biology experiments was performed with the nonparametric Wilcoxon test, unpaired data significant levels <0.02.

Total cell lysates were obtained with hot (95°C) lysis buffer (125 mmol/L Tris HCl pH 6.8, 2.5% SDS), separated on SDS-PAGE gradient gels (Invitrogen) and transferred to nitrocellulose membrane. Densitometry of immunoblots was performed using ImageJ (https://rsb.info.nih.gov/ij/): phospho-AKT and phospho-GSK3β signal were divided by the intensity of the corresponding total protein. GLUT1, phospho-EGFR, EGFR, phospho-MET, and MET were normalized over HSP90 or GAPDH.

Multiplex phosphoproteomic analyses were as in ref. 25 with commercially available plates [MAP kinase multiplex assay (K15101D) and phosphoSer(473)AKT/total (K15100D)] MesoScale Discovery. Phospho-proteins signals, normalized over control, were converted in logarithmic form and represented with heat maps generated by the Gedas program. Measurements of endocytosis of surface biotinylated receptors with sulfo-NHS-SS-biotin (ThermoScientific) and assessment of surface GLUT1 with noncleavable biotin were performed by capture ELISA as in ref. 25.

Immunofluorescence was performed as previously described (26). Primary antibodies were revealed by Alexa Fluor 555- and 488-conjugated antibodies. In the GLUT1/EEA1 staining, permeabilization was in PBS 2% BSA, 0.1% saponin for 30 minutes at room temperature. Incubation with primary and secondary antibodies was in PBS 0.2% BSA, 0.1% saponin. In the phospho-AKT/AKT total stainings, blocking and permeabilization was in PBS 5% donkey serum, 0.3% Triton-X100. Primary and secondary antibodies were diluted in PBS 0.3% Triton-X100. In the phospho-EGFR/EGFR total stainings, blocking and permeabilization were in PBS 2% BSA, 0.02% saponin for 10 minutes at room temperature. Confocal analysis was performed with a Leica SP8 AOBS microscope. Colocalization between GLUT1 and EEA1 was measured with JacoP plugin of ImageJ. Because variability was recorded among experiments in total-AKT staining, the intensity of phospho-AKT was measured as the Pearson correlation coefficient between phospho-AKT and Phalloidin provided by the LasAF software (Leica). Measurements of myc-GLUT1 endocytosis are detailed in figure legend.

Cells were counted by Trypan blue exclusion. Briefly, cells silenced or treated as indicated were collected both from the medium and by trypsinizing the adherent ones. This was necessary because some of the breast cancer cell lines growth either partially in suspension (HCC1187) or detach easily from adherence. Cells were resuspended in complete medium, diluted in Trypan blue solution and counted with a Bürker chamber. At least eight fields/sample were counted in multiple replicates. In the glucose deprivation assays, cells were starved in DMEM 10% FBS without D-glucose and sodium pyruvate (A14430 Gibco).

Cell toxicity and ATP cell content were evaluated with the CellTox Green Citotoxicity Assay and CellTiter-Glo (Promega).

Glucose uptake was measured with the Glucose Uptake-Glo Assay (Promega). Cells were incubated for 10 minutes in glucose-deprived medium supplemented with 1 mmol/L 2-Deoxy-D-Glucose (2DG). Glucose uptake (fmol/cell/min) was calculated as follows:

Intracellular lactate production was determined with the fluorimetric L-Lactate Assay Kit (Cayman). The same number of HCC70 and HCC1187 silenced cells were processed and the concentration (μmol/L) of L-Lactate was determined as follows:

CF is the corrected fluorescence (background subtracted), and slope and intercept were calculated from the standard reference curve.

We used the cBioPortal (27, 28) to survey alterations affecting USP6NL in breast cancer datasets. In the METABRIC cohort (23, 29), high mRNA and gene amplification of USP6NL were detected in ∼7% and ∼3.5%, respectively (Fig. 1A, left). Around 75% of all amplified cases also displayed mRNA upregulation (Fig. 1A, left). Moreover, breast cancers with gains and copy number amplifications showed a proportional increase in USP6NL mRNA expression (Fig. 1B). Similar results were obtained from The Cancer Genome Atlas dataset (Fig. 1A, right; ref. 30).

In the METABRIC cohort, expression of USP6NL was higher in the PAM50 Basal and Her2 subtypes versus luminal and normal subtypes (P < 0.0001; Fig. 1C). Using the genomic-based integrative cluster classification of breast cancers (31), we found that USP6NL expression was associated with the Integrative Cluster 10 (IntClust10; Supplementary Fig. S1), which incorporates mostly triple-negative tumors and represents a high-risk group.

In univariate analysis, USP6NL copy number and expression was significantly associated with worse breast cancer–specific survival (HR 1.81; P < 0.0001; Fig. 1D). This association remained significant in multivariable analysis (HR* 1.46; P < 0.004) when grade, size, and lymph nodes involvement were considered (Supplementary Table S1).

We investigated the expression of USP6NL in breast cancers also at the protein level by IHC on TMAs (Materials and Methods and Fig. 1E). The normal mammary gland displayed heterogeneous staining of USP6NL, with levels comparatively higher in the luminal than in the basal layer (Fig. 1E). USP6NL protein expression was, on average, higher in breast cancer samples (P < 0.0001). Strongest positivity (score 3) was found in 7% of breast cancers and never detected in normal samples (Fig. 1F), concordant with the upregulation at the mRNA level (Fig. 1A). Also in this cohort, strong USP6NL expression correlated with aggressiveness (higher grade and increased percentage of Ki-67 positive cells) and with the molecular subtype of triple-negative breast cancers (Supplementary Table S2). In the IHC series we did not detect correlation with prognostic outcome likely because of a lower number of events in the USP6NL-high breast cancers in this cohort versus the METABRIC one. Altogether, these results suggest that USP6NL is a copy number driver gene in breast cancers.

To investigate the role of USP6NL elevation in breast cancers, we selected two USP6NL-high basal-like breast cancer cell lines expressing high amount of the USP6NL protein, HCC70 and HCC1187, and two nontransformed, USP6NL-low, mammary cell lines, MCF10A and HMEC (Fig. 2A). Silencing of USP6NL reduced proliferation in both the USP6NL-high lines, leaving the USP6NL-low cells unaffected (Fig. 2A and B; Supplementary Fig. S2A and S2B), arguing that USP6NL-high breast cancer cells depend on this molecule for their optimal growth. The effects of the functional ablation of USP6NL on cell number were due to a decrease in the proliferation rate and not to increased cell death, as it will be subsequently shown.

USP6NL has been found to restrain EGFR endocytosis (17–19). Endocytosis and trafficking control the magnitude of the EGFR signaling (32), suggesting that USP6NL might affect proliferation by dampening intracellular signaling pathways. Thus, we analyzed alterations in the phosphorylation status of key signal transducers: AKT, ERK1/2, JNK, and p38, by the MesoScale platform. USP6NL depletion significantly reduced AKT phosphorylation in HCC70 and HCC1187 cells, but not in MCF10A (Fig. 2C; Supplementary Figs. S2B and S3A). Lesser changes in ERK1/2, JNK, and p38 phosphorylation were also detected, albeit frequently not significant and not shared between the USP6NL-overexpressing breast cancer cells (Fig. 2C).

To validate the specificity of the effects, we performed rescue experiments. We generated stable inducible cell populations by infecting HCC70 cells with lentiviral vectors carrying the silencing-resistant version of USP6NL or its GAP-defective mutant R150 (19). Re-expression of the wild-type protein in USP6NL-silenced cells rescued AKT activation and cell proliferation (Fig. 2D and E). Conversely, re-expression of R150 did not recover any of the phenotypes, revealing that the effects of USP6NL depend on its GAP function (Fig. 2D and E).

Phosphorylation of AKT by PDK1 on Thr308 and by mTORC2 on Ser473 results in full AKT activation (33). USP6NL silencing similarly impaired both phosphorylation events, and the phosphorylation of the AKT substrate GSK3β, in the overexpressing breast cancer lines but not in MCF10A cells (Supplementary Fig. S3A).

Finally, expression of the myristoylated, constitutive active, PI3K mutant (PI3KCA; ref. 34) recovered AKT activation in USP6NL-silenced cells (Fig. 2F). PI3KCA also rescued the defect caused by USP6NL depletion on cell proliferation (Fig. 2G), suggesting that down modulation of the PI3K/AKT axis might account for this latter effect in the USP6NL-silenced breast cancer cells.

Because the GAP function of USP6NL inhibits EGFR endocytosis (17, 19) and controls AKT activation (Fig. 2D), we tested whether reduced phosphorylation of AKT in USP6NL-silenced cells resulted from increased internalization and degradation of the EGFR. Depletion of USP6NL in the breast cancer-overexpressing cell lines, but not in MCF10A or HMEC cells, accelerated EGFR degradation dampening both EGFR and AKT phosphorylation both at steady-state (Fig. 3A) and upon EGF stimulation (Supplementary Fig. S3B). Measurements of EGFR endocytosis confirmed that functional ablation of USP6NL in HCC70 cells increased the amount of internalized EGFR (Supplementary Fig. S3C). Conversely, another RTK, the MET receptor, was largely unaffected by USP6NL silencing, suggesting some degree of specificity of USP6NL, within the limits of the tested conditions. In particular, total MET levels (Fig. 3A) and receptor internalization (Supplementary Fig. S3C) were not appreciably affected. Some, not significant, increase was detected in the levels of phospho-MET, not correlating with AKT phosphorylation or USP6NL levels. Together these results are compatible with the notion that overexpression of USP6NL delays EGFR endocytosis and degradation, thereby chronically activating AKT.

A likely downstream candidate for the effects of USP6NL in the described phenotypes is RAB5. USP6NL exerts GAP activity toward several Rabs (17–20), among these, RAB5 has been directly implicated in endocytic pathways and specifically in EGFR endocytosis (35, 36). We tested, therefore, whether overexpression of RAB5 could phenocopy the ablation of USP6NL. This was indeed the case, as shown by reduced EGFR levels, and AKT phosphorylation (Fig. 3B), implicating RAB5 as a probable downstream effector of USP6NL in the control of the analyzed phenotypes.

AKT controls cell proliferation, survival, and metabolism (33). Thus, we looked at the ATP cell content and at the occurrence of cell death in the USP6NL-silenced cell lines. In these experiments, we introduced also a lung carcinoma cell line, H1299, which has little amount of USP6NL, to test the effects in a transformed background bearing low USP6NL levels (Supplementary Fig. S4A and S4B). Depletion of USP6NL reduced the ATP cellular content in HCC70 and HCC1187, but not in USP6NL-low cells (Fig. 4A) without causing significant differences in cell death, as measured by DNA accessibility (Fig. 4B), suggesting a role for USP6NL in the metabolic control of cancer cell growth.

One of the best-characterized metabolic outcomes of the upregulation of AKT in cancer is increased aerobic glycolysis (37). In USP6NL-depleted breast cancer lines, but not in USP6NL-low MCF10A and H1299 cells, we found a ∼80% decrease in the rate of glucose uptake (Fig. 5A). Similarly, lactate production was significantly reduced (Fig. 5B). Finally, only the USP6NL-overexpressing breast cancer cell lines were sensitive to glucose deprivation (Fig. 5C). Notably, USP6NL depletion abrogated this phenotype (Fig. 5C), indicating that the elevated glycolytic metabolism of these cells depends on the high USP6NL levels.

AKT stimulates the glycolytic capacity of cancer cells by increasing expression and targeting of GLUT1 to the PM, and by regulating key glycolytic enzymes (14–16, 37). In particular, AKT stabilizes GLUT1 at the PM by inhibiting its endocytosis (12, 13). In agreement, we found that USP6NL depletion decreased the amount of GLUT1 in the overexpressing breast cancer cell lines, but not in the USP6NL-low cells where, rather, an increment in the glucose transporter was observed (Fig. 6A). Re-expression of USP6NL, but not of R150, in the silenced HCC70 cells readily recovered GLUT1 levels showing that the latter depend on the GAP activity, hence—most likely—on the endocytic/trafficking function of USP6NL (Fig. 6B). In further support, treatment with the lysosomal blocking agent bafilomycin, but not with the proteasomal inhibitor MG132, rescued the levels of GLUT1 in USP6NL-silenced cells, indicating that USP6NL stabilizes GLUT1 by inhibiting its trafficking-mediated degradation (Fig. 6C). Finally, and in agreement with a GAP-dependent role for USP6NL in GLUT1 endocytosis and degradation, also overexpression of RAB5 reduced GLUT1 levels (Fig. 6D).

We analyzed whether the effect of USP6NL on GLUT1 stability are mediated by AKT. The expression of PI3KCA in the USP6NL-silenced HCC70 cells, along with the rescue of AKT phosphorylation (Fig. 2F), also recovered GLUT1 expression (Fig. 6E). In addition, treatment with the AKT inhibitor MK2206 resulted in GLUT1 degradation, which was rescued by concomitant addition of bafilomycin, exclusively in the USP6NL-overexpressing breast cancer cells (Fig. 6F). Induction of GLUT1 degradation in the HCC1187 cells required high inhibitor concentration. However, MK2206 did not cause GLUT1 degradation in the USP6NL-low cells, even at the highest concentration (Fig. 6F). Altogether these data argue that increased USP6NL levels stabilize GLUT1 in an AKT-dependent manner.

Next, we tried to obtain evidence linking the stabilization of GLUT1 in USP6NL-overexpressing cells to AKT-dependent inhibition of endocytosis of the transporter. We analyzed the dynamics of GLUT1 transport upon EGF stimulation in HMEC cells overexpressing USP6NL (USP6NL^high) or its GAP mutant (R150^high; Fig. 7A; Supplementary Fig. S5A). In control HMEC, in the absence of stimulation (SF), GLUT1 mainly localized in intracellular compartments, partially overlapping with the early endosomal marker EEA1. Stimulation with EGF for 5 minutes caused significant redistribution to the PM and reduced colocalization with EEA1. When EGF was removed to interrupt stimulation, GLUT1 relocalized to early endosomes. Similar dynamics were observed in HMEC-R150^high cells. Conversely, in HMEC-USP6NL^high, GLUT1 was predominantly at the PM under all conditions, including in serum starved cells, suggesting that increased USP6NL levels sustain basal AKT activation that, in turn, maintains GLUT1 at the cell surface. In support, we observed basal activation of AKT in serum starved USP6NL^high cells (Supplementary Fig. S6A and S6B). Moreover, consistent with data showing that AKT blockade prevents localization of glucose transporters to the PM (38), pretreatment with MK2206 prompted GLUT1 removal from the surface in control and R150^high cells stimulated with EGF as well as in the serum-starved USP6NL^high cells (Fig. 7B; Supplementary Fig. S5A). The ability of USP6NL to interfere with GLUT1 endocytic dynamics was also confirmed by two additional approaches exploiting either a GLUT1 construct carrying an exofacial myc tag or surface biotinylation experiments (Supplementary Fig. S5B and S5C). Finally, similarly to USP6NL depletion, inhibition of AKT by MK2206 treatment impaired proliferation in the breast cancer cell lines overexpressing USP6NL (Supplementary Fig. S7).

In conclusion, our findings argue that high USP6NL levels stabilize GLUT1 in an AKT-dependent manner fostering aerobic glycolysis, and determine sensitivity to glucose deprivation in breast cancer cells.

Our study reveals a previously unrecognized crosstalk between endocytosis and metabolic reprogramming in breast cancer relying on the overexpression of USP6NL. This event is likely to play a causal role in the natural history of breast cancer, as USP6NL is frequently amplified/overexpressed in this cancer, in particular in basal-like tumors classified as IntClust10, and correlates with worse prognostic outcome. In addition, in vitro, the depletion of USP6NL impairs proliferation exclusively of breast cancer cells overexpressing this protein and not of normal mammary cells. This suggests a degree of addiction of USP6NL-overexpressing breast cancers, pointing to this alteration as a relevant mechanism of signaling-metabolic rewiring.

Under basal conditions in breast cancer cells that overexpress USP6NL, functional depletion of the latter promotes EGFR and AKT downregulation. We hypothesize that chronic USP6NL overexpression, by delaying EGFR endocytosis (probably by inhibiting the action of RAB5), stabilizes the receptor at the PM facilitating its activation and the ensuing recruitment of PI3K, eventually resulting in sustained AKT phosphorylation. This hypothesis is supported by the observation that ectopic expression of USP6NL in HMEC cells increases AKT phosphorylation even in absence of exogenous EGFR stimulation. Furthermore, USP6NL is mainly localized to the PM (19), a location where the PI3K/AKT signaling axis is generally considered to initiate (39). Finally, AKT phosphorylates the GLUT1 endocytic adaptor TXNIP at the PM, inhibiting the endocytosis of glucose transporters (40). Thus, the concerted action of USP6NL and AKT, acting at the cell surface, might link EGFR signaling to glucose uptake.

Similarly to the EGFR, also GLUT1 degradation is accelerated by USP6NL depletion, and by AKT inhibition through MK2206, only in USP6NL-overexpressing breast cancer cells. Interestingly, MK2206 is an allosteric AKT inhibitor that prevents insulin-driven localization of GLUT1 to the PM and glucose uptake (38) but does not cause GLUT1 degradation in normal cells (this study and ref. 38). In agreement, only cell lines overexpressing USP6NL are sensitive to AKT inhibition and glucose starvation. Of note, depletion of USP6NL abrogates sensitivity to glucose deprivation, in the same cell lines, further indicating that the high glycolytic capacity of these cells relies on USP6NL.

Based on the sum of our results, it is tempting to speculate that the poor prognostic outcome of breast cancer patients carrying high USP6NL copy number might be related to the ability of USP6NL to promote glycolysis. Indeed, elevation of glycolysis is generally associated with tumor aggressiveness and poor prognosis, and is a distinguishing feature of a subset of basal-like breast cancers (2, 3). Interestingly, in many instances, upregulation of glycolysis in tumors occurs through overexpression/stabilization of GLUT1, an alteration that correlates with aggressiveness (41–45). In this context, our data argue that the amplification/overexpression of USP6NL might be a mechanism of metabolic adaptation/fitness for some breast cancers. If so, USP6NL overexpression might represent a point of metabolic vulnerability for a subset of aggressive basal-like breast tumors.

No potential conflicts of interest were disclosed.

Conception and design: D. Avanzato, E. Pupo, L. Lanzetti

Development of methodology: D. Avanzato, E. Pupo, A. Sapino

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Avanzato, E. Pupo, N. Ducano, G. Bertalot, C. Luise, S. Pece, C. Caldas, P. Paolo Di Fiore, A. Sapino

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Avanzato, E. Pupo, C. Isella, C. Luise, S. Pece, A. Bruna, O.M. Rueda, C. Caldas, P. Paolo Di Fiore, L. Lanzetti

Writing, review, and/or revision of the manuscript: D. Avanzato, C. Caldas, P. Paolo Di Fiore, A. Sapino, L. Lanzetti

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Luise

Study supervision: C. Caldas, L. Lanzetti

We thank Miriam Martini and Emilio Hirsch for helpful discussion and for reagents, Carlos Sebastian for suggestions and for critically reading the manuscript. Davide Disalvatore for bioinformatics support. Work in the authors' lab is supported by grants from the Associazione Italiana per la Ricerca sul Cancro (AIRC Investigator Grant, project 15180 to L. Lanzetti and 14404, MCO 10.000 and extension to P.P. Di Fiore; 5xmille MCO Extension to A. Sapino), Fondo Ricerca Locale 2017 (University of Turin) to L. Lanzetti, MIUR (the Italian Ministry of University and Scientific Research) to P.P. Di Fiore, the Monzino Foundation to P.P. Di Fiore, FPRC 5xmille Ministero Salute 2011, 2014 to A. Sapino and 2015 to L. Lanzetti, and Cancer UK to A. Bruna, O.M. Rueda, and C. Caldas.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.